Re: INTC / IRQ Numbers

[Kristoffer E. <kri...@gm...>]

Big thanks to you and Markus.

I'll pull down a fresh lethal.git tree and see where it fails, and push the fix upstreams.
Whats broken in general?

Best wishes
Kristoffer

On Thu, 23 Aug 2007 18:53:40 +0900
"Magnus Damm" <mag...@gm...> wrote:

> Hi Kristoffer,
> 
> On 8/24/07, Kristoffer Ericson <kri...@gm...> wrote:
> > Greetings,
> >
> > Didn't get a reply on my last mail, so here goes again. Basicly how do I calculate the IRQ number using the INTC table in setup-sh770x.c.
> 
> Hehe, your email are still on my todo list.
> 
> I wanted to answer your question, but I hesitated a bit because i'm
> still not really sure why the official tree is broken and what the
> state is of your tree. I wish we could work together and fix up the
> official 2.6.24 tree step by step.
> 
> > For instance I want to know where IRQ4 ends up, If I check how the RTC IRQ's are calculated and use the I end up at IRQ 30.
> > Not sure If I understand it properly.
> 
> There is a private set of enum_ids for each set of tables passed to
> the intc code using register_intc_controller(). The enum_id value zero
> needs to be treated as unused and all other values are used to provide
> mappings for the intc code between interrupts and bits or bit fields
> in registers. For setup-sh770x.c the first useful enum_id is IRQ0.
> These enum_ids are private data for the intc code and there is no
> point in trying to use them somewhere else.
> 
> What you are looking for is hidden inside vectors[] - the information
> provided by INTC_VECT(). The first argument to INTC_VECT() is the
> private enum_id and the second one the vector number specified by the
> processor data sheet. This vector number is hard coded in the
> processor chip and the intc tables are used to describe the
> relationship between these vectors and certain registers.
> 
> You should look at the evt2irq() macro in include/asm-sh/irq.h to
> figure out the relationship between the processor vector number
> (second argument to INTC_VECT()) and the IRQ number passed to
> request_irq(). The PINT07 interrupt source for example is located on
> vector 0x700 and using the evt2irq() macro this gives us IRQ 40.
> 
> Hope this helps. I'd be happy to help out hacking up the new pint
> controller but I feel we need to get the upstream kernel in a better
> shape first. Hint: send patches! =)
> 
> Happy hacking!
> 
> / magnus


-- 
Kristoffer Ericson <Kri...@Gm...>